Admin panels
CRUD over production tables with row-level rules, search, bulk actions and an audit trail, in an afternoon.
For developers
Install the ToolJet plugin and your agent builds real ToolJet apps: pages, queries, components and permissions, against your actual data sources. Drop into JavaScript or Python when the visual layer is not enough. Nothing you build is a throwaway.
Build a refund approvals app on the orders table in our Postgres. Support agents request, finance approves over $500, every decision logged.
list_datasourcesfound prod-postgres (read/write), stripeget_datasource_query_schemaorders, refunds, userscreate_appRefund approvals · 2 pages · 7 componentsadd_querieslistPending, approveRefund, logDecisionmanage_app_permissionssupport: request · finance: approvelint_app_spec0 errors · 1 warning (empty state)App created in development. Open in the editor or promote to staging.
The ToolJet plugin
The plugin exposes ToolJet's own contracts to your coding agent: data source schemas, component catalogue, query lifecycles and permissions. The build workflow can validate planned app changes before applying them, then check the saved result. It runs on your agent's subscription, not ToolJet AI credits.
# 1. Create a personal access token in ToolJet (Profile Settings → Personal access tokens)
# 2. Add the ToolJet MCP server to your agent: docs.tooljet.ai/docs/build-with-ai/mcp/setup/
# 3. Ask for an app
> Build an on-call rota app on the engineers table in prod-postgres
list_datasources → prod-postgres, pagerduty
create_app → On-call rota · 2 pages · 9 components
manage_app_permissions → platform-eng: edit · everyone: view
lint_app_spec → 0 errorsTalks to what you already run
The tools that never make the roadmap, without a React project each time.
CRUD over production tables with row-level rules, search, bulk actions and an audit trail, in an afternoon.
Multi-step approvals with role-gated actions, Slack or email notifications and every decision recorded.
Live views over Postgres, Snowflake or your APIs, with drill-downs and actions wired to queries.
Customer lookups joining your database, Stripe and Zendesk, with guarded write actions.
On Self-hosted, build webhook, scheduled or app-triggered workflows with JavaScript logic, data-source nodes and run logs.
Bring your own React component for the one widget the catalogue does not cover; everything else stays declarative.
From your agent through the plugin, or on the canvas. Both write to the same application model, so a teammate can pick up either way.
JavaScript and Python queries in apps, with custom JavaScript logic in Self-hosted workflows. Custom components for the rest.
Development, staging and production environments with environment-scoped credentials, Git sync and version history with rollback.
Apps inherit workspace SSO and roles. Set page, component, query and row-level access as part of the build, not after.
In production
Built in, not bolted on
Authentication, authorisation, secrets, environments and audit come with the platform, so an internal tool does not need its own auth middleware or deployment pipeline.
SAML, OIDC and LDAP with directory groups mapped to roles and SCIM provisioning.
Workspace, app, page, component and query-level access, plus row-level security where the data needs it.
Data source secrets never reach the browser or the agent; queries run through ToolJet with environment-scoped credentials.
Git sync, versioned releases, rollback and separate environments. Agent changes wait in development until you promote them.
Security and governance features vary by plan. Workflows and Agent Builder are available on Self-hosted only; BYOK is a Self-hosted Enterprise add-on. Compare plan availability.
Two ways to get there
Self-serve
Start free. Describe the tool, refine it on the canvas, and ship it under your own access rules. Start with one app, then test it with your team before rolling it out.
Start freeProfessional services
When the timeline is fixed or the process is tangled, our engineers build alongside you: scoped delivery, production-ready apps, and knowledge transfer so your team owns the result.
Talk about a projectThe agent builds against ToolJet's application model, not free-form code: components, queries, events and permissions your team can open in the visual editor afterwards. You skip auth, hosting, secrets and deployment for each tool, and the result is maintainable by whoever is on the team next year.
Yes. JavaScript and Python queries in apps, with custom JavaScript logic in Self-hosted workflows. You can also build custom React components for anything the catalogue does not cover. Code is the escape hatch, not the starting point.
The agent can inspect schemas and run queries through ToolJet when the connecting user has permission. Query results can therefore be visible to the agent and its model provider. Data source credentials stay in ToolJet. Use a workspace token with appropriate access and review queries before running them.
Apps move through development, staging and production with environment-specific credentials. Git sync keeps app definitions in your repository, and version history lets you roll back. Agent-built changes land in development.
Yes. Docker, Kubernetes, Helm and OpenShift, in your VPC, on-premise or air-gapped. The Community Edition is open source under AGPL-3.0; the MCP server is MIT-licensed.
Install the ToolJet plugin, point it at your workspace and describe the app. Or start on the canvas, free.
Page updated